Combined cluster and discriminant analysis: An efficient chemometric approach in diesel fuel characterization Márton Novák a , Dóra Palya a , Zsolt Bodai a , Zoltán Nyiri a , Norbert Magyar b , József Kovács c , Zsuzsanna Eke a,d, * a Eötvös Loránd University, Joint Research and Training Laboratory on Separation Techniques (EKOL), 1/A, Pázmány Péter sétány, Budapest 1117, Hungary b Budapest Business School, University of Applied Sciences, Department of Methodology, 9-11, Alkotmány utca, Budapest 1054, Hungary c Eötvös Loránd University, Department of Physical and Applied Geology, 1/C, Pázmány Péter sétány, Budapest 1117, Hungary d Wessling International Research and Educational Center, 56, Fóti út, Budapest 1047, Hungary A R T I C L E I N F O Article history: Received 26 July 2016 Received in revised form 13 November 2016 Accepted 16 November 2016 Available online 23 November 2016 Keywords: Combined cluster and discriminant analysis Chemometrics Diesel fuel Compound-specific isotope analysis Environmental forensics A B S T R A C T Combined cluster and discriminant analysis (CCDA) as a chemometric tool in compound specific isotope analysis of diesel fuels was studied. The stable carbon isotope ratios (d 13 C) of n-alkanes in diesel fuel can be used to characterize or differentiate diesels originating from different sources. We investigated 25 diesel fuel samples representing 20 different brands. The samples were collected from 25 different service stations in 11 European countries over a 2 year period. The n-alkane fraction of diesel fuels was separated using solid-state urea clathrate formation combined with silica gel fractionation. The stable carbon isotope ratios of C10–C24 n-alkanes were measured with gas chromatography–isotope ratio mass spectrometry (GC–IRMS) using perdeuterated n-alkanes as internal standards. Beside the 25 samples one additional diesel fuel was prepared and measured three times to get totally homogenous samples in order to test the performance of our analytical and statistical routine. Stable isotope ratio data were evaluated with hierarchical cluster analysis (HCA), principal component analysis (PCA) and CCDA. CCDA combines two multivariate data analysis methods hierarchical cluster analysis with linear discriminant analysis (LDA). The main idea behind CCDA is to compare the goodness of preconceived (based on the sample origins) and random groupings. In CCDA all the samples were compared pairwise. The results for the parallel sample preparations showed that the analytical procedure does not have any significant effect on the d 13 C values of n-alkanes. The three parallels proved to be totally homogenous with CCDA. HCA and PCA can be useful tools when the examining of the relationship among several samples is in question. However, these two techniques cannot be always decisive on the origin of similar samples. The initial hypothesis that all diesel fuel samples are considered chemically unique was verified by CCDA. The main advantage of CCDA is that it gives an objective index number about the level of similarity among the investigated samples. Thus the application of CCDA supplemented by the traditionally used multivariate methods greatly improves the efficiency of statistical analysis in the CSIA of diesel fuel samples. © 2016 Elsevier Ireland Ltd. All rights reserved. 1. Introduction Middle distillate fuel oils such as diesel fuel are frequently spilled in the environment. Those cases when the source of the spills is questionable or totally unknown are providing serious challenges in environmental forensic investigations. In order to determine the liability associated with the cleanup and remedia- tion chemical fingerprinting methods are applied. In the case of source correlation studies of unknown fuel contaminations the evaluation of similarities or dissimilarities among samples or among a sample and a possible source is the problem to be solved. * Corresponding author at: Eötvös Loránd University, Joint Research and Training Laboratory on Separation Techniques (EKOL), 1/A, Pázmány Péter sétány, Budapest 1117, Hungary. E-mail addresses: marton.novak@ekol.chem.elte.hu (M. Novák), dora.palya@ekol.chem.elte.hu (D. Palya), zsolt.bodai@ekol.chem.elte.hu (Z. Bodai), zoltan.nyiri@ekol.chem.elte.hu (Z. Nyiri), Magyar.Norbert@uni-bge.hu (N. Magyar), kevesolt@geology.elte.hu (J. Kovács), eke.zsuzsanna@wirec.eu (Z. Eke). http://dx.doi.org/10.1016/j.forsciint.2016.11.025 0379-0738/© 2016 Elsevier Ireland Ltd. All rights reserved. Forensic Science International 270 (2017) 61–69 Contents lists available at ScienceDirect Forensic Science International journal homepage: www.elsevier.com/locat e/f orsciint